I want to find contributions of each independent variables

Question

I would like to get contributions of independent variables (factors driving sales) in absolute dollar values.

The log linear equation looks like this:

$$\ln(\text{sales})=b_1 \times \ln(\text{baseline})+b_2 \times \text{weather} +b_3 \times \text{holiday} +b_4 \times \text{promos} + ... $$ I want a model that can tell me something like this:

In a \$40M sales turnover, weather contributes \$4M, holiday contributes $1M, etc..."

Can anyone suggest a method that should be followed here?

Is there a particular reason that the model's beta values are not enough with a given sales figure? — Upper_Case, Jan 10 '17 at 17:19
you have a relative (growth rate) model, and want absolute contribution. this can't be done. — Aksakal, May 30 '19 at 14:15

score 0 · Answer 1 · answered May 30 '19 at 14:10

It can't be done directly - lets exponentiate the equation:

$sales = e^{b_1 \times ln(baseline)} \cdot e^{b_2 \times weather} \cdot e^{b_3 \times holiday}$

We can determine the individual terms, but they do not add. At best we can say: Over the $10M baseline weather provides 8 times increase while holidays divide by two (0.5 times increase)

I want to find contributions of each independent variables

1 Answers1